NaN in posterior density

Estimation
1

Hi! I estimated a model and everything seems good (trace plot, mode check plots and identification), however, there is a shock standard deviation that has no posterior (SE_eps_ir):

Active_16
Active_16875×656 43.4 KB

When I looked at oo_.posterior_density.shocks_std.eps_ir only NaN’s are shown.
Would anyone help me to figure it out what is happening here?

Thank you

2

Can you upload the content of the metropolis-folder?

3

Thanks for your time, prof Pfeifer.
Here they are:

Files.zip (2.2 MB)

4

Looking at the _mh*_blck-files the standard deviation of that parameter is in the range of 1e-11. That is, the posterior is almost degenerate. It seems the problem for that parameter already appears in the mode_check-plot

5

Thanks for your response, professor Pfeifer.

  1. I could not find the standard deviation of the SE_eps_ir _mh*_blck-files. I guess it was in x2 matrix, right? But I did not understand what it means.

  2. Looking at the mode_check plot, I thought the graph was strange because the x-axis is in a very small range. Should the posterior mode be at the maximum of the posterior likelihood even in a mode_check plot with a very tiny x-axis range?

  3. Finally, what is the solution for this problem generally speaking? My prior for the standard deviation of eps_ir are already unlimited, so I do not see a immediate solution.